Proof of Type Input MCP server
Commit Oct 11, 2026Enables AI agents to integrate human-verified text input components that prove content was typed by a person rather than generated by AI.
Also known as yogev77/proofoftype-input, proofoftype-input
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 1.3.0
- Last commit
- Oct 11, 2026
- License
- MIT
What is Proof of Type Input MCP?
Proof of Type Input enables AI agents to integrate human-verified text input components that prove content was typed by a person rather than generated by AI.
Summary generated from the project's README
About Proof of Type Input MCP server
Proof of Type provides a way for developers to ensure that text submitted to their websites, such as comments or reviews, was actually typed by a human rather than generated by an AI. Instead of relying on unreliable AI detectors, this service uses a secure input method that tracks typing rhythm and requires mobile confirmation via Face ID or fingerprint. This creates a signed proof token that confirms the content was entered letter by letter without pasting.
Developers can integrate this into their workflow using an MCP server with tools like get_quickstart, verify_proof, and explain. The server can be added to clients like Claude, Cursor, or VS Code via a remote HTTP transport. Once implemented, the system provides an EdDSA-signed JWT that includes metadata like a human typing score and the author's username. This allows for local or remote verification to ensure the text matches the original input and originated from the correct site.
What you can do with it
- 1Add human-verified comment boxesAsk an AI agent to implement a comment section that requires proof of human typing.
- 2Verify text authenticity on the serverValidate that a submitted comment was typed by a person and matches the provided token.
- 3Prevent AI-generated spamUse hardware-backed mobile signatures to stop automated scripts from posting content.
Written from the project's README. Every use case is checked against a quote from it.
How to install Proof of Type Input MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall Proof of Type Input MCP in Claude Code
$ claude mcp add --transport http proofoftype-input https://www.proofoftype.com/mcpInstall Proof of Type Input MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"proofoftype-input": {
"url": "https://www.proofoftype.com/mcp"
}
}
}Install Proof of Type Input MCP in VS Code
.vscode/mcp.json
{
"servers": {
"proofoftype-input": {
"type": "http",
"url": "https://www.proofoftype.com/mcp"
}
}
}Install Proof of Type Input MCP in VS Code CLI
$ code --add-mcp '{"name":"proofoftype-input","type":"http","url":"https://www.proofoftype.com/mcp"}'Install Proof of Type Input MCP in Codex
$ codex mcp add proofoftype-input --url https://www.proofoftype.com/mcpInstall Proof of Type Input MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.proofoftype-input] url = "https://www.proofoftype.com/mcp"
Install Proof of Type Input MCP in Gemini CLI
$ gemini mcp add -s user --transport http proofoftype-input https://www.proofoftype.com/mcpInstall Proof of Type Input MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"proofoftype-input": {
"url": "https://www.proofoftype.com/mcp",
"type": "http"
}
}
}Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableyogev77/proofoftype-input is public and reachablegithub.com/yogev77/proofoftype-input ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-11github.com/yogev77/proofoftype-input ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeinitialize answered with protocol 2025-06-18 in 916 msPass
README
View on GitHubProof of Type Input
A text box that proves a human typed it, not AI.
Add one script tag, and the comments, posts, reviews or articles on your site come with proof that a real person typed them: signed in to Proof of Type, typed letter by letter (no paste, no AI-generated text), and confirmed on their phone with Face ID or a fingerprint. Your server gets the text plus a signed proof token that you, or anyone, can verify.
- Live demo and docs: https://www.proofoftype.com/developers
- API reference: https://www.proofoftype.com/developers/api
- For AI assistants: https://www.proofoftype.com/llms.txt
<script src="https://www.proofoftype.com/sdk/v1.js" async></script>
<form method="post" action="/comments">
<div data-proofoftype data-name="comment" data-placeholder="Write a comment"></div>
<button>Post</button>
</form>
When the form is submitted, the person confirms on their phone, then your server receives:
| Field | What |
|---|---|
comment |
the text, exactly as typed |
comment_proof |
the proof token (an EdDSA-signed JWT) |
Build it with your AI coding agent (MCP)
Vibe coding a site? Connect the Proof of Type MCP server and ask your agent for "a comment box that proves a human typed it".
claude mcp add --transport http proofoftype https://www.proofoftype.com/mcp
Other clients (Cursor, VS Code, …):
{ "mcpServers": { "proofoftype": { "url": "https://www.proofoftype.com/mcp" } } }
Tools: get_quickstart(framework), verify_proof(token, text, origin), explain(topic).
Why
AI-generated text is now free and endless, and AI detectors can only guess from the finished words, so they're wrong both ways. Proof of Type doesn't guess. It proves how the text was made:
- The person types in our input, key by key. It runs in a secure frame from proofoftype.com, so the host page, extensions or scripts can't fake keystrokes.
- The typing is timed on our server (it can't be instant or replayed) and scored for human rhythm.
- The person approves the exact text on their phone in the Proof of Type app, with Face ID or a fingerprint. The phone's hardware vouches for the app (Apple App Attest, Android key attestation). A script can't do this step.
Pasted or generated text never gets a proof.
Verify on your server
Always check the proof before you save the text. Accept it only if it's valid, matches the text, and was made on your site.
const r = await fetch('https://www.proofoftype.com/api/v1/verify', {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
token: req.body.comment_proof,
text: req.body.comment, // the exact text
origin: 'https://your-site.com', // your site
max_age_seconds: 600, // made in the last 10 minutes
}),
}).then((res) => res.json());
if (!r.valid || !r.text_matches || !r.origin_matches || !r.fresh) {
return res.status(400).send('Please type your comment in the box.');
}
// Use each proof once: store r.proof.id with the comment, refuse repeats
// (proof tokens are public — they're in proof links).
// r.proof.author: the Proof of Type username who typed it
// r.proof.link: a link pasted with the text, if any
Prefer a helper? Copy server.mjs into your project (verifyProof({ token, text, origin })).
Or call the API directly (no key needed):
curl -s https://www.proofoftype.com/api/v1/verify \
-H 'Content-Type: application/json' \
-d '{"token":"<comment_proof>","text":"<comment>","origin":"https://your-site.com"}'
{ "valid": true, "text_matches": true, "origin_matches": true,
"proof": { "id": "pot_…", "author": "yogev", "origin": "https://your-site.com",
"typed_at": "2026-10-11T18:20:31.000Z", "score": 87, "words": 6, "text_sha256": "…" } }
Offline verification
Proof tokens are JWTs signed with EdDSA (Ed25519). Public keys: https://www.proofoftype.com/.well-known/jwks.json.
import { createRemoteJWKSet, jwtVerify } from 'jose';
import { createHash } from 'node:crypto';
const JWKS = createRemoteJWKSet(new URL('https://www.proofoftype.com/.well-known/jwks.json'));
const { payload } = await jwtVerify(token, JWKS, {
issuer: 'https://www.proofoftype.com',
audience: 'https://your-site.com',
});
if (payload.text_sha256 !== createHash('sha256').update(text, 'utf8').digest('hex')) throw new Error('Not the signed text');
| Claim | What |
|---|---|
iss |
https://www.proofoftype.com |
sub |
proof id (pot_…) |
aud |
the site it was typed on |
iat |
when it was verified |
typed_by |
"human" |
author |
Proof of Type username |
signed_on |
"ios" or "android" |
text_sha256 |
SHA-256 (hex) of the exact UTF-8 text |
score, wpm, words, chars |
human score (0–100), typing speed, length |
JavaScript API
const input = ProofOfType.mount('#editor', {
placeholder: 'Write your article…',
maxWords: 2000,
button: false, // verify from your own button
});
input.on('change', ({ words }) => {});
const proof = await input.verify(); // { text, html, token, author, score, url, … }
Options: name, placeholder, maxWords (10–3000), theme (auto | light | dark), button, buttonLabel, form, typewriter (show the carriage animation, default true), typewriterColor (e.g. '#e11d48'), sound (default true), allowLinks (default true: one pasted link is signed in the proof, read it as proof.link from the verify response; false blocks all pasting). result.html is display formatting only and isn't covered by the proof: sanitize it, or use result.text. The same options work as data- attributes (data-max-words, data-button-label, …).
Events: ready, signin, change, verified, error, resize, also dispatched on the element as DOM events (proofoftype:verified).
React
Copy react.jsx into your project. It loads the SDK for you:
import { HumanInput } from './react.jsx';
<HumanInput placeholder="Write a comment" onVerified={(r) => save(r.text, r.token)} />
TypeScript types: index.d.ts.
Privacy
- Proof of Type never stores the text typed on your site. Only its SHA-256 is signed into the proof.
- Keystrokes are checked and discarded. The text reaches your page only once it's verified, and only your page's real origin.
- Your site gets the person's Proof of Type username, never their email or account.
- Each proof is single-use: one typing session, one phone signature, one proof.
What's open source
This repository: the browser SDK, the React component, the server helper, types and examples (MIT). The checks themselves (keystroke capture, typing timeline, phone signatures) run on Proof of Type, where the host page can't change them. That's what makes the proof worth something.
FAQ
How do I know a comment on my site was written by a human and not AI? Use this input as the text box and verify the proof on your server. A proof only exists when a signed-in person typed the text in the input and approved it on their phone.
Is this an AI detector? No. It proves the process (typed, timed, signed by a person) instead of guessing from the text.
Do my users need an account? Yes: a free Proof of Type account and the Proof of Type app (iPhone or Android). Signing in is one popup; each proof is one Face ID or fingerprint.
Can people paste? Only links. Autocorrect of typed words is allowed; predictive text, swipe and dictation are not.
What does it cost? Free for humans.
Links
- Docs: https://www.proofoftype.com/developers
- API: https://www.proofoftype.com/developers/api
- OpenAPI: https://www.proofoftype.com/openapi.json
- Changelog: https://www.proofoftype.com/developers/changelog
- Proof of Type: https://www.proofoftype.com
MIT © Proof of Type
Proof of Type Input MCP questions
How do I add Proof of Type Input MCP to Claude Code?
Run claude mcp add --transport http proofoftype-input https://www.proofoftype.com/mcp in your terminal. Add --scope user to make it available in every project.
How do I add Proof of Type Input MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
How do I know a comment on my site was written by a human and not AI?
Use this input as the text box and verify the proof on your server. A proof only exists when a signed-in person typed the text in the input and approved it on their phone.
Is this an AI detector?
No, it proves the process of typing, timing, and signing by a person instead of attempting to guess based on the finished text.
Do my users need an account?
Yes, users require a free Proof of Type account and the Proof of Type app on iPhone or Android to approve proofs via Face ID or fingerprint.
Can people paste text?
Users can only paste links; other forms of pasting, predictive text, swipe, and dictation are not allowed.
What does it cost?
The service is free for humans.